Senior Backend Engineer, Agent Infrastructure

Airbyte is the data and action layer for AI agents. We give agents fast, accurate, authenticated access to business data across hundreds of sources, so they can discover the entities that matter, reason over real-time context, and take action in the systems they read from, not just observe them.

We started as the open-source standard for data movement and proved the economics of data integration at scale: hundreds of connectors, thousands of companies, and, since 2020, have raised $181M from leading investors including Benchmark, Accel, Altimeter, Coatue, and Y Combinator. As our CEO Michel Tricot puts it, "the last ten years were all about structured data. The future is all about context." We're now building that context infrastructure for production-grade agents on the same open foundation, as agents become the primary consumers of enterprise data.

Our mission is unchanged: make data available and actionable to everyone, everywhere. That everyone now includes AI agents.

The Role

Airbyte is building the runtime that powers production-grade AI agents.

This is not an ML research or prompt engineering role. We’re looking for an experienced backend or platform engineer who has built complex production systems and has also worked hands-on with AI agents, orchestration, retrieval, or LLM-powered applications.

Our AI Runtime sits between language models and the enterprise systems agents need to understand and act on. It is responsible for assembling the right context, identifying entities across systems, selecting tools and connectors, coordinating execution, validating evidence, enforcing permissions, and handling failures safely.

A user may ask something simple like:

“Investigate this customer and tell me what changed.”

Behind that request, the runtime may need to identify the customer across multiple systems, retrieve fresh data, determine which tools to call, coordinate multiple steps, validate the result, and return an answer with evidence the user can trust.

That infrastructure is what you’ll help build.

We’re still early. The architecture is taking shape, but there is significant room to influence how the runtime is designed, how developers interact with it, and how we make agentic systems reliable in production.

What You’ll Do

Build the AI Runtime

Design and implement the orchestration layer that turns natural-language intent into reliable, multi-step execution.

Build systems for routing, state management, retries, failure handling, and long-running agent workflows.

Develop systems for entity resolution, context assembly, connector orchestration, and evidence retrieval.

Build reusable Skills that encapsulate business workflows and domain-specific capabilities.

Coordinate connectors, APIs, tools, deterministic logic, retrieval systems, and multiple language models.

Build Reliable Agent Infrastructure

Design systems that retrieve and assemble the right enterprise context at runtime.

Build evidence-backed reasoning with citations and traceability that users can inspect and verify.

Implement permission models, freshness validation, and action policies for production environments.

Build evaluation, replay, observability, and debugging systems that help us understand why agents succeed or fail.